179 / Teresa Torres: Is AI Re-Prioritizing Delivery Over Discovery (Again)?

product community seemed like we were in this cycle where we were building the wrong stuff. When you were on the podcast, you described what product discovery was and how to do it. What has changed in the last four or five years? I think one of the challenges with AI is it's pulling us in the wrong direction. And I'm a huge fan of AI. We can get into that. This is not an anti-AI sentiment. It's that AI is making it easier and faster to build. And so we're once again putting even more emphasis on delivery and we're forgetting about, hey, maybe we should be asking, is this the right thing to build? How do they figure out how to get going with a good, you know, kind of continuous discovery model in a world where now, you know, we're not just learning those fundamentals, but we're trying to apply those fundamentals with all of these new tech approaches coming in, you know, especially with AI? I actually think that discovery becomes more important. Big surprise. We see a lot of like half-baked AI features that don't really work, right? And, you know, there's a period of time where you're going to get a little bit of lift from saying AI this and AI that. But that period of time actually might already be over. I think we're starting to see some backlash against AI. We're starting to see some backlash against the Foundation Labs and what kind of world they're creating for us. And I think... it's no longer going to be good enough to just slap on some AI features. How do we get to that point where we can have organizations recognize that we're sprinting in the wrong direction? And how can we bring them back into, I think one of the core tenants of product discovery is... is mapping opportunities, right? So how do we get back to ensuring that everybody's going after the right opportunities? I wish I had to answer. I will tell you, I feel like I've been advocating for discovery for 20 years and we've made progress, but there's still plenty of companies and plenty of teams that don't know very much about their customers and don't talk to them and still think they can sit in a room and have all the answers. And that's going to get a lot worse. Dan, we just got done with Teresa Torres. She was amazing. She's been on the podcast before, but we had an all new experience this time. Tell us about it. Well, first off, had to not be starstruck. Continuous Discovery Habits was one of the seminal books I read when learning how to be a good product manager. Our discussion was awesome, right? We talked about AI and how it is relating into continuous discovery skills, the importance of maintaining those skills. And then we went all the way to just talking about, you know, understanding business and good product habits. Amazing. Amazing stats about just the number of people that aren't able to connect what they're doing with providing value to the business. This is a great episode. Let's get after it. We are here today with Teresa Torres. Teresa Torres is an author, a prominent voice and writer in the product space and a past guest of the Product Momentum podcast. Episode 58 in April 2021. And we're super excited to have her back. Her book, Continuous Delivery Habits, is an essentially required reading for product people. So, Teresa, welcome back. Thanks. It sounds like the last time I was here was literally right before the book came out. Yes. The book came out in May of 2021. Yeah. And I apologize for correcting you, but it is Continuous Discovery Habits. Just so folks can find it. Thank you. So, when you were here last. You talked about how at the time we were under emphasizing discovery. We were over emphasizing delivery. We seemed like we were in this. And when I say we, I mean like the product community seemed like we were in this cycle where we were building the wrong stuff. When you were on the podcast, you described what product discovery was and how to do it. What has changed in the last four or five years? You know, I feel like we were on this trajectory probably since like let's say 2005 to 2020 ish maybe 22 i'm going to put it at the release of chat gpt where we were actually getting better companies were starting to recognize we got to include the customer in the process we need faster feedback loops i'm not going to say we were getting great at it but we were i would say steadily getting better i think one of the challenges with ai is it's pulling us in the wrong direction And I'm a huge fan of AI. We can get into that. This is not an anti-AI sentiment. It's that AI is making it easier and faster to build. And so we're once again putting even more emphasis on delivery. And we're forgetting about, hey, maybe we should be asking, is this the right thing to build? And there's parts of this I love. Like we're seeing some companies, like I always like to look at the companies that are on the bleeding edge because I feel like it's a little hint of what our future might be. And we do have companies where The marketing team is shipping to production. The sales team is shipping to production. This is both awesome and terrifying. And the awesome part is that like in the same way that like we want to empower product teams and get as many people close to the customer, like impacting the product, that's amazing. It's terrifying because who's creating product coherence and how do we make sure we're serving the market? Now, I think there's ways to enable this and allow both. But I think over the last three years, we're starting to see a negative trend on the discovery side, which definitely worries me. But I'm hoping this exact same technology is going to help us course correct and help us do better discovery. How do we lean into those teams that actually want, you said like, you know, marketing, shipping, shipping to production, right? Yeah. These other stakeholders are clearly excited. They've got now they have the tools to do it. How do you still leverage and how do we make sure they feel like they're a part, an important part of the process of getting the right thing built without maybe giving them like the driver's seat all the time? You know, here's what I love about what's happening with AI. I feel like humans are makers. Some of us happen to be on a path where we learn to maker skills and we really are makers. Some of us, maybe those, what was required to learn those skills was intimidating. But I feel like we're still makers. Like humans are inherently creative. We can envision a future. And I think AI is like unlocking this ability for more people to be makers and more people to be builders. And I actually think that's going to be net positive in the long run. I think as companies, we have to learn new skills to allow everybody to contribute, but still end up with a coherent product. And there's a company I just read about. I can't remember. It's one of the new, like super sexy, fast growing startups, like a ramp or a linear. I don't remember which one it was. I apologize. Maybe I can find it afterwards and we can put it in the show notes. But what they do is they literally let anybody push to production, but gated. So they have a pool of like lab customers, customers that have opted in to saying, give me anything and everything you ever create. They have about 5,000 customers in that pool. And literally anybody in the company can push and release to that pool. But then they have really strict controls on what goes from that pool into the general audience. And I actually really like that as a gating mechanism. This is really similar to Facebook's harness where they say any engineer can push to 1%. But then we're really rigorous about what goes from 1% to 100%. And I think what I like about this is it allows us to collect real data on what works. And we, for a long time, have said anybody can have a good idea. And this really sort of puts the rubber to the road, right? Like we're letting anybody push to this group of people that have said we want to try anything and everything. But then we're being really disciplined about how we're basically saying for our earliest adopters, we're going to be super innovative and give them anything and everything. that we throw at them. But for everybody else in the market, we're going to be really thoughtful and careful and make sure that we maintain a coherent product. And I have a feeling we're going to see more of that. I think that's roughly in broad strokes the right model. Is that like, sorry, I think it seems like with that model, it's almost replacing like, you know, Hi5 mockup type user testing. Is that what you're seeing where it's like? you're almost taking concept level features, getting in front of the real users and learning that way, as opposed to like a pre-production type discovery model? It depends. It's going to depend on a lot of things. First of all, 5,000 customers is still a lot of people. And so I might want to make sure that I'm doing some usability testing. I'm doing some prototyping even before I push to that population. It's going to depend on the nature of the change. Right? Like some changes probably don't need that. Some changes might. It's going to depend on whether what we're touching involves money. Right? If we're looking at payments and like it's got to be rock solid. Like you don't want to mess with payments for 5,000 customers. You don't want to mess with payments for five customers. So it's still kind of, we still have to tailor our discovery to the risk involved. And I think these types of mechanisms allow us to manage a lot of risk, but still not all of it. I don't think prototyping is going away. I don't think like just concept testing with customers is going away, but I think we're going to get to faster, interactive things that feel like real products much quicker. And I think that's a net positive. Yeah. You mentioned how they're letting, yeah, kind of everyone send a product. I feel like I just heard a whole bunch of dev managers just scream in terror. Yeah, it's, there's, like I said, there's a lot about this that's really terrifying. Even like as a, like the designer in me is like, who's making sure we have a coherent user interface? And like, how do we know, like one of the things I hate is as companies grow and especially companies bought by private equity where they're like squeezing the orange, they'll add feature after feature after feature after feature and none of them work well together. And as a consumer, this is like my biggest pet peeve of like, okay well i'm using feature a and feature b and therefore i can't use feature c because it doesn't really work with feature a and feature b and it's like i'm responsible for figuring out what's the combination of features that work well whereas to me that's the product team's job and why aren't they doing that and it's because this is like a classic symptom of a feature factory and this is where i think like ai is going to make us worse at this like but but we're going to quickly get better at it um and i think like AI kind of breaks all of our norms about like how fast we can go and how much we can produce. And so I think we're going to have to develop new ways of, um, testing for coherence and, and like inner compatibility of all these different features. Um, but the part I love is let's unlock the creativity of everybody in our organization and let them contribute. Yeah. A minute ago, you were mentioning companies on the bleeding edge. Um, yeah. And I'm wondering, you know, Obviously, discovery hasn't been solved everywhere. What about those companies, especially maybe ones that don't see themselves as traditional tech companies and software companies that still obviously software is a huge part of their business? How do they figure out how to get going with a good kind of continuous discovery model in a world where now we're not just learning those fundamentals, but we're trying to apply those fundamentals with all of these new attack approaches coming in yeah with yeah especially with ai yeah i actually think that discovery becomes more important big surprise um we see a lot of like half-baked ai features that don't really work right and you know there's a period of time where you're going to get a little bit of lift from saying ai this and ai that But that period of time actually might already be over. I think we're starting to see some backlash against AI. We're starting to see some backlash against the foundation labs and what kind of world they're creating for us. And I think it's no longer going to be good enough to just slap on some AI features. So I think we're going to see even our traditional companies, our airlines, our banks, whatever, our retail companies. there's lots of opportunity to use this technology well and to solve real customer problems. In fact, since we've talked, I've started a podcast called Just Not Possible where I interview product teams about the AI features they're building. And there's some pretty amazing stories out there. So for example, I just interviewed a team. They're a travel company. They book travel for employees. So their ideas are trying to make... business travel as easy as possible. If you book through this company, you don't have to fill out expense reports. Everything's paid for. They're just trying to remove this bureaucratic work of paying on a card, getting a receipt, filling an expense report, having it be approved. And one of the challenges they ran into in their business was they would reserve... hotels with a virtual credit card. And some hotels struggle with virtual credit cards. And so the employee would show up and their hotel wasn't actually booked because the payment didn't go through. And so their customer care team had to call, make like 10,000 calls a week to hotels to make sure payment actually was processed correctly. Well, they built a voice agent to do this. And they now have a voice agent that is making 10,000 phone calls a week. And they're actually verifying at a higher rate than their human team was able to do because they can make more calls. And this error that they were encountering where someone would show up in their hotel room wasn't actually booked has come way down. And they're saving a ton of money because they're not having humans do it. And they still have a customer care team and they still all have their jobs. They're just going back to actually taking care of their customers instead of like calling hotels and saying, did you charge my card correctly? And that's such a great use case of like a very real customer problem with costs involved. that they've completely solved. And I think there's going to be plenty of opportunity for this in our traditional businesses. But that's not a very sexy product, right? That's not something they're splashing on their homepage saying, look at this amazing AI thing we built. But it is something that just created a ton of business value and works really well. And I think that's just for you, Sean. Yeah, I actually was going to say that I think that somewhere the ITX travel desk is hearing this and is like, we got to get this for Sean because I'm not a schedule and budget and booking guy. That's probably the lower end of my lower end of my overall skill set. You said that we're going to experience some pain early because we are going to build. We are going to be because of the ability to build the wrong thing faster. We are going to build the wrong thing faster. How do we get teams to where do you think the breaking point is going to be? How do we get to that point where we can have organizations recognize that we're sprinting in the wrong direction? And how can we bring them back into the I think one of the core tenants of product discovery is is is mapping opportunities, right? So how do how do we get back to ensuring that everybody's going after the right opportunities? Yeah, I wish I had to answer. I will tell you, I feel like I've been advocating for discovery for 20 years and we've made progress, but there's still plenty of companies and plenty of teams that don't know very much about their customers and don't talk to them and still think they can sit in a room and have all the answers. And that's going to get a lot worse. In fact, I'll tell you this past week, Anthropic just announced their Anthropic interviewer where they basically had Claude interview 1,250 people. I actually find this to be a really compelling use case. When you use Claude, you get a little pop-up that says you have 15 minutes to participate in an interview. You can do it via text, you can do it via voice, and Claude is literally interviewing you. Well, Anthropic made all the transcripts available, and I started looking through them. What's funny is Anthropic reported they had their user researchers involved in this process, but if I look at the questions that Claude asked, they're not good interview questions. And so, like, I actually was just having a conversation with Claude right before this podcast trying to check some of my own assumptions. I was like, when is this type of question a good question? Like, am I wrong in thinking this is not a good question? We went back and forth and Claude even provided research that supports. And I was like, I framed it as like, tell me when this is a good question. Right? Like, I'm trying to, like, disprove my bias. I think, like... The challenge we have, and again, I've been trying to advocate for this for 20 years. Most people, including people that have research in their job title, they learned on the job. They learned methods that are not grounded in actual research methods. And we're seeing a ton of really poor quality research out there. And I don't, I don't know why this is. Like we have good books. We have good people out there teaching this stuff. We have amazing thought leaders in the space. Like go read Rob Fitzpatrick, go read Indy Young, go read Steve Portigal, read my book. Like there's tons of people out there teaching how to conduct a good interview, but most people are still doing the wrong thing. And I think like what this comes and even like big companies with user researchers are still doing the wrong thing. And I think part of this is who do we hire as researchers? We hire people out of school with PhDs. okay if you have a phd in anything other than let's say hci or human factors you didn't necessarily know how to do industry research like if you have a phd in anthropology you probably did one large scale research project in a specific context and it's up to you to translate that to industry i'm not criticizing that person that's their research and their experience or we have researchers that came into industry I'm making a lot of enemies right now with this rant, but we're going to just push through it. Or we have people that came into industry and they learned from their peers, but their peers didn't necessarily learn about like good, valid, reliable research questions. And so I think what's hard is companies start to do good, start to do what they think is good discovery and they don't get high quality output from it because they're not asking the right questions. They're not doing valid, reliable research. And I don't, uh to be honest i don't know how to fix this in fact before we started recording i said i'd rather talk about ai than discovery and it's because i kind of have discovery fatigue like i'm running out of ways to try to teach this and try to spread these ideas because i don't i've been doing it for so long and like there is a a microcosm of people that are like really engaging and learning this but there's still a vast majority that don't and so i think what's gonna happen the reason why i think this is gonna get a lot worse is we're like Anthropic, we're going to use AI to interview our customers. We're going to ask bad questions at scale. And then we're going to do bad synthesis at scale. And we're going to still build what we always wanted to build in the first place. And I know this sounds super cynical. I wish I was maybe a little happier this morning about this. But that is where we are. Like that is our reality. And so I think what's going to happen is we're going to have like... really good companies with good researchers that are asking thoughtful questions. They're gonna learn how to do that at scale with AI. They're gonna learn how to do AI synthesis at scale at a high quality. And hopefully what it means is they're gonna run ahead and we're gonna start to learn from them. Because if I've learned anything, it's that the way companies change is they see a really sexy, successful company do something and they try to adopt that company's practice. So hopefully we'll see some companies run ahead and be way better than everybody else. And the way they got there was by doing better research and it will help spread the practice. But I think it's going to get ugly before it gets better. Now, the father of this podcast, Sean Flaherty, he does a lot of different talks, but one of the sections of his talk, he talks about kind of like the cycle of grief in software. And I feel like potentially we're in this, we could be in this like cycle of grief. in the way you can explain is that the bottom is going to be the bottom is going to be when they realize how much bad investment has been made right well we see new things share that like so how do we shrink that like how do we shrink that and get to like the part where there there is some positive return from the pain that we pain that we experience but i think what you're saying right now is you feel like we might be we might just be driving down into the bottom of the cycle then hopefully we're on the upswing back up soon yeah i also will just share like that's a really that change curve and the and the like the dip of grief like is it's probably where i'm at with discovery personally if it wasn't just clear from my previous answer like you have i do this for so long and like not see change to be like okay well i probably i probably should stop pushing this boulder up this hill um and that's partly why like i've been geeking out on ai so much so i've been doing a ton in my own personal work and i've been starting to build tools because i actually am starting to think maybe instead of training humans how to do discovery well I should be building tools for AI to do discovery well. Because I think what we're seeing from businesses over and over and over again is that for humans, it feels like we're right and we can sit in a room and have the right answer. And so we don't make the decision to invest in discovery. We don't introduce doubt. We just think there's no time. And so maybe the right answer is let's build discovery tools for AI to do that for you and for AI to do it well. I don't know. Maybe. That's my current half-baked hypothesis. I love it. I mean, we're product people, right? Most of our hypotheses are half-baked. Yeah. Yeah, kind of going back to what you're saying and having, you know, folks maybe come and discover whose backgrounds are in, you know, broader, you know, anthropological, that was a $10 word, you know, having that kind of background versus, you know, maybe product managers with industry knowledge. Do you see when you talk to folks that maybe the folks with the industry knowledge can be more easily trained in asking good interview questions versus vice versa? Or are you just like, you know what, AI, I'm going to teach AI how to ask the right questions here. You know what, I think both people with both those backgrounds bring different skills and are very valuable. So if I have a PhD in anthropology and I did like a long period qualitative research study and i learned ethnographic methods and i learned how to do good observations that's incredibly valuable in industry what you have to learn when you come to industry is that industry doesn't work at the pace of academia and so you have to learn how do you translate those research methods to be good enough for the pace of industry and that's where i see phd researchers really struggle is they talk about like Why do we have to go so fast? Why can't we slow down and get it right? If we do it right the first time, we'll save time in the long run. No business leader is ever going to buy that argument. And I think this is why a lot of researchers struggle when they come into business. The flip side is you learned in industry, you actually work at the pace of industry, but you have zero concept of reliability and validity and what leads to good research. I actually think both can learn this, right? Like we've taught... since 2017 i've had a course on a very simple format of interviewing which is just story based interviewing what i like about it is it's reliable and it's valid it's good research methods it's fairly simple to teach i can explain the concept to you and you're gonna get it what's hard is learning to do it well and it takes a lot of practice and so i think people don't realize they're not doing it well and that they need more practice. And this is why the very first AI tool I built was an interview coach, because I want to start exposing the gap between how good you think you are and where you're actually at so that I can motivate you to do a little bit more practice. And so I don't think the problem is people's backgrounds. Like I think generally people... are doing the best they can and i'm not criticizing even like i'm really curious to talk to an anthropic researcher that was involved in this project because like i want to get their point of view like maybe there's something i'm missing about maybe they have different research goals than what i'm assuming and that question was perfectly valid i'm struggling to see what the research goal would be that would lead for that being a valid question but i'm genuinely curious and so like this is what's hard like i don't ever do product teardowns because if someone on the outside I don't know what constraints they're working with. I don't know who they're trying to serve and who they're leaving out and like what decisions they're making. And from the outside, I can't know that. And I think that's true with research too. Like I bet if I had a conversation with the anthropic researcher, they might have a very valid reason for why they did it that way. And maybe this was just a proof of concept. They were just trying to see, can we do this? And they weren't really worried about the quality of their questions. So like, I don't, I don't want to criticize. I can envision a world. where like literally every product team is collecting really rich customer stories and that leads to better products. I really want to create that world. I'm sort of struggling with what's the best way to do that because I feel like I've been doing this for a long time. And maybe I have to just embrace the like, I don't know if you guys have heard the like PhD saying of like you come into a phd thinking you're going to change the world and you leave realizing if you have a wildly successful career you might put a little tiny debt in the bubble uh maybe i just need to embrace like get over my own ego and realize like the amount of change i've already impacted is enough and to stop worrying about the rest uh but i don't know i don't know how to i'm i'm running out of ideas What I was going to say is I think it's the putting it into practice part that's really hard. I think people get these ideas. I think discovery is becoming more popular. I think the challenge is that when you try to do good discovery, you're kind of swimming upstream against how business wants to work. Business still, a large percentage of people in business leadership positions still have this really strong belief that to do business well. we sit in a room, we come up with a strategy and we execute on the strategy. And so like once you introduce this like, yeah, but you need a feedback loop from your customers, that's a little jarring. And I think as long as that's true, it's going to be really hard to get teams doing this well for a long, long time. And I think that's the like change part that I struggle with. Yeah, you know, and with the discovery, I feel like it's the same thing that you run into even with a competitor analysis, right? Like you don't, with the research, you don't really know what hypotheses there after you see the outputs. It's like, hey, do it with the competitors, but like you don't have the same constraints or knowledge of kind of what they faced when they built the thing. They're talking about, you know, the decision making, the speed. One of the tools that you've brought kind of the forefront, I know it's been done elsewhere, but the idea of the opportunity mapping and opportunity solution trees. I would love to hear how that's changing in the world of AI. We were just interviewing recently for the pod with Phil Hornby and we were talking about, as product managers, one of our key jobs is actually making decisions. And then how do we speed up making good decisions? So yeah, I'd love to kind of hear where you're going with your opportunity solution tree model. You know, when I developed the opportunity solution tree, what I was looking for was what's the underlying structure of discovery that doesn't change. And I think I nailed it. So I don't, I don't expect the, I don't expect the, that's a, that sounds like a very egotistical thing to say. I don't mean it in workshops. So you're allowed to say that. I don't mean like it's perfect and it's never going to change. I just mean like my goal was. Discovery is really messy. We have a lot of tactics. You learn a lot and learning is always messy and there's twists and turns and you might think you're going to go from A to B, but as you go along the way, you discover C is a better destination. And I think what I was trying to do with the opportunity solution tree was humans have a variable capacity to deal with messiness. how do we guide a cross-functional team that probably has variation in their ability to deal with messiness and help them understand where they are and what they're doing? And so with the opportunity solution tree, all I was trying to do was to understand, like, as a team, we should have a clear definition of what success looks like. That's our outcome. We need to understand who our customers are. That's the opportunity space. And we need to understand how what we're building is serving our customers and serving our business. And that Like when I describe it that way, like do I think that's ever going to change? I don't like we're always going to have to create value for our business. We're always going to have to create value for our customer. Hopefully we're doing one thing to do both of those things and not like doing competing things. And so I think the like visual is not likely to change. I was really trying to find what's the simplest underlying. structure that gives us like a mental representation that as we do all the messy work, we can attach things to. Now, what I do think is changing is maybe, I'm going to say maybe, we're going to find new inputs for how to understand the opportunity space. And then I think we're definitely going to find new ways to discover solutions. So with the opportunity space, the reason why I say maybe, I've historically said opportunity should emerge from customer interviews. And lots of teams say, well, why can't they emerge from sales conversations? And why can't they emerge from sales tickets? I mean, support tickets. And why can't they emerge from behavioral analytics? And I've often said, because those other sources are missing context. When a support ticket comes in, I don't know what the customer was trying to do when they encountered that problem. and I need to go collect that full story before I can fully satisfy that need. Okay, it is now possible, and this is the exciting part about what Anthropic did, it is now possible, I believe, to conduct interviews at scale. So if I can take what Anthropic did and turn Claude into a better interviewer, and I could interview 1,500 people in a week, and we figure out how to do good synthesis on that, which is also a hard problem, but I do believe this is now possible. If we can interview at scale, what is creating the opportunity space look like? I'm not sure. I think that is going to evolve and change. That's exciting to me. Like I'm excited to see what does this new technology enable? But I think it's a huge open question. Like one of the benefits of the team doing the opportunity mapping is they build and doing the interviewing themselves is they build empathy. They get a lot of exposure to their customer. I don't want to lose that. But we could augment it with interviewing at scale, which I find exciting. And then I think in the solution space, we're already seeing things change radically. The fact that I can go to Lovable or Bolt or Replet and create an interactive prototype with no engineering skills whatsoever. I can assumption test way faster. I can assumption test with higher fidelity. I can test multiple concepts with higher fidelity much faster. So I think we're seeing the mechanics, the... around how we populate our tree change. But I don't think, I will be surprised if the structure of the tree changes anytime soon. One of the things that we like about the Opportunity Tree is it definitely gets people to somewhere where they can determine there's something that a user wants that they don't currently have right now. Yeah. Where it actually challenge, where we find the biggest challenge is how do they convert that to something that will benefit the business? You like tactics just I love this conversation because this is like this is like this is like bare product like core product tactics, right? How do how do we get from we've done this analysis? We know that the user will want we know the user wants this to help them solve a problem. How do we get them to really understand how that benefits the business? Yeah, so one of the challenges is the opportunity space is infinite. If I just go start interviewing customers about what they want. I'm going to find a lot of things that don't necessarily create business value. And this is why the root of the tree is your outcome. And your outcome represents business value. Even if you should translate a business outcome to a product outcome, we can talk about that. But it still represents business value. And then your outcome constrains the opportunity space. You're not going to go talk to customers about anything and everything they could possibly need. you're defining your research questions and your interview question in the context of your outcome. So as you explore things and you uncover opportunities, you're using that outcome as a filter. So you're only addressing the opportunities that you think will create value for that business. By definition, will create value for the customer, but you're only choosing the ones that will create value for the business. Now, Sean, to your question, like how do we get teams to actually do this? What's hard? Most companies aren't using outcomes well. Their outcomes are glorified outputs. They're not tied. I like to define outcomes in the context of your products revenue model. Like how do we create value for our business? We bring in more revenue. Like that's just like business 101, right? So we have to understand how does our product make revenue? What are the inputs to that formula? Like what are the variables in that formula? What are the inputs to each of those variables? So let me walk through an example to make this really tangible. If I run a subscription business, my revenue formula is going to be the number of subscribers I have times how much they pay each month times how long, how many months they stick around, right? That is my revenue model formula. Now for each of those variables, what are the inputs? How do I acquire customers? Well, I have to attract them to my website. I have to convert them into paying customers. How do I increase how much they spend every month? Well, I probably have tiered subscriptions. So I have to look at how to upsell them over time to a higher tier. So that's looking at engagement and engagement with premium features and the desirability of those features, how well they deliver on the value they promise. And then I want to look at lifetime value. And again, that's tied to reducing churn. driving engagement over time. So I can look at the inputs to those. And then eventually I want to start to look at, okay, what are the behaviors in the product that are leading indicators of these business outcomes? And because that's what my product teams can influence. My product teams can't just grow revenue unless you're an e-commerce company, right? My product teams influence customer behavior in the product. So now we're looking for what are the customer behaviors in the product that if we increased them would increase engagement. and therefore would increase retention or would increase visitor to customer conversion rate and therefore would help me acquire more customers. And a lot of companies are not this disciplined about how they set outcomes, right? They either let their product teams just pick a random metric, which I don't understand that at all, or the product leader is just communicating a business outcome. Our goal this quarter is to grow revenue by 10%. And they're not doing the work to like help their teams translate this to like, what are the behaviors in the product that actually move this number? And product teams don't know how to do that. I can tell you that right now. They have no idea how their work connects to revenue. And so I think to do this well, leaders have to fill in that gap. I asked the question because that what you just mentioned about people. product teams really not understanding how the work that they're doing is actually supporting the revenue model. It's something that we see either with new people that are coming into the agency or with clients that we're working with. That's like one of the biggest things that we help educate from, from a product perspective. We also definitely, we run our own workshops at ITX and like we use the opportunity tree model. Dan actually brought, Dan was actually instrumental in bringing that into our workshop model because it's so it's so effective and it helps draw that it helps really draw that conclusion for people yeah we do uh um every other year we do a continuous discovery how that's benchmark survey and we've had 2600 people fill out the survey uh so fairly large scale a full 20 could not answer the question about whether their outcome was intended to drive revenue or to reduce costs like 20 of people said i don't know that's shocking to me Yeah. So to your point, like how do we get product teams driving like business outcomes, creating value for the business? The first thing we need to do is teach them the business context, right? Like if they can't connect the dots between their work and whether or not it drives revenue and reduce costs, we have a problem. They're not going to create value for the business. This is something I think what's hard is a lot of people in leadership positions either went to business school and they learned this in business school. or they learned it on the job and they forgot that they learned it on the job. They assume everybody knows it. And I'll tell you, I became a startup CEO at a pretty young age. I was an employee and then I became the CEO of the startup I was an employee at. So I wasn't a founder. And I remember when I stepped into the CEO role, actually before that, I was the VP of operations. It was the first time I saw the books. It was the first time I was exposed to any finance at any company whatsoever. And I remember talking to my CFO and doing like, okay, hold on a second. Revenue recognition is different from cash flow. What is this pretend world you have over here of like the rules of revenue and recognize revenue and it's not just money came in the bank and then money goes out the bank. Like that was foreign to me. I had no idea. I had zero exposure to accounting and how finance works, right? This is true for our teams, but business leaders throw things around like bookings and billings and revenue recognition and your product teams have no idea what you're talking about. like none whatsoever. But leaders are like so steeped in this world, they don't realize that. So there's this huge gulf between the language of business and the language of product teams. And we have to do a better job of filling that gap for sure. Oh, that's so true. I remember the first time that like, yeah, I was working on a software project and you're talking about like depreciable assets and like software being a depreciable asset. Yeah, I went to engineering school. It's like, I'm going to have to go ahead and Google that because nothing that you said made sense to me. Okay, now I understand. I've capitalized budget, expense budget, but most people come in a product. You're not versed in that level of finance and accounting practices. Yeah. You have to have... Forget even the finance part of it. We meet teams that they don't... I run monthly community calls with my community. And then we have this woman, super bright, very good product manager. She started to have challenges with her head of marketing. And so we started talking about like, how about just go talk to your head of marketing and learn about what are their goals? Like what's their quarterly goals? She didn't know. So she had a meeting where they talked it through and she learned things like, oh, our customers pay for the software that I build. I thought it was just a value add. I didn't even know customers paid for it. This is a very capable, strong product manager. Something about her business context meant she had no idea that her customers actually paid for her software. This is shocking to me. Is it her fault? Absolutely not. I will tell you, I've known her long enough to know she's a very capable product manager. We forget to close the gap. How is your product sold? How does your customer pay for it? How does that create value over time? What's the value creation moment in the product that leads them to continue to renew? Like if our product teams don't know this, they can't create business value. But we uncovered this in our first survey. Our first survey, I believe, was in 2022. And since then, Hope Gurion, one of my business partners, and I have been recording a ton of content. Like if you go to Product Talk... We have an in-depth guide on setting outcomes and as part of that guide, we have a video on understanding if your outcome drives revenue or reduce costs. We talk about the most common mistakes teams make with outcomes. We're actually launching a course in January. I'm not sure when this episode is going to come out, but in January we're launching a course called Business Fundamentals. where we're just, we hope created this business fundamentals canvas where it just gives your product teams like 14 questions to answer about the business side of your product, just trying to fill in the gaps. Because we were really surprised how big this gap exists. And it makes it really hard for product teams to deliver value if you don't close the gap. Yeah, you can confirm to those resources are awesome. Like I've checked it now for the outcome setting. Awesome. Super helpful for you. You know, getting the top of your opportunity tree, like having some sense of what you're doing and what you're thinking about. This has been an amazing episode. Thank you so much for being here today. I know the listeners don't know, but you and I are both suffering from cold. So we made it. We didn't really like cough throughout the entire, I think we coughed throughout the entire episode. So congratulations. I have a couple of quick takeaways and then we're going to ask you a fun question at the end. So one of the big takeaways AI is bringing us back to kind of like this overemphasis on delivery. And we as product people need to ensure that we are causing the matter of creating product coherence. I liked your quote that humans are makers, right? We've talked about builders a lot on the episode on other episodes, but I kind of like that concept of humans are makers better. A big one, don't mess with payments. Don't mess. Don't mess with payment, right? Like you said, even if it's only five people, don't mess with payments, right? Learn to ask thoughtful questions at scale and understand that people that are in your organization will have varying levels of capability with messiness, but that learning is messily and valuable. So I kind of like tying those two concepts together. And then where we kind of like wrapped up was learn the language of business. and then educate your teams sean i love that you found really positive takeaways despite my little side rants about research and discovery and being in that gulf trough of a little bit of despair yeah dan will dan will tell you i am i am the i am the eternal optimist i am that guy that will always find some i will always find the the rosy outcome right you know what's funny is i'm usually the eternal optimist and i'm still optimistic about discovery i i think that like I'm working through my own like messy reaction to AI driven discovery could be really powerful. And it also can completely cannibalize everything that's valuable about discovery. And so like, I'm like literally in that moment of recognizing no matter how much I want humans to do interviewing, as soon as interviewing is available at scale by AI, no company is going to do human interviewing. Like that's probably our reality. And so I think I'm just like, wrestling with how do I feel about that and how much do I embrace that? And it's put me a little bit in that trough of despair. So hopefully this wasn't a downer. Like I am optimistic. I just, we're living in a weird time. Yeah. And it's December and it's dark all the time. And it's December and it's dark all the time. The last big takeaway, which I'm going to replace our normal last question with is, can you tell us everything about the Just Now Possible pod and where we can listen to it? Yeah, okay. I started building my first AI product this past March. And as I was building it, I was like, wow, I just feel like I'm learning this all by myself. And there's some, you know, okay resources out there on the web, but I just want lots of stories about what product teams are really doing and how are they learning how to do this. And it feels like the 90s. We're all in the frontier. It's like the web is new again, but this time it's AI. And so what I really wanted to do was interview people about how they're figuring this out and what they're building. And I love it because it has really allowed me to embrace my cross-functional background and like nerd out with engineers and nerd out with data scientists and nerd out with product managers and nerd out with designers. So I interviewed the full cross-functional team that is building an AI product. We cover everything from what customer problem are you solving? How did you find that problem? How did you prototype to figure out if AI would work? to literally the AI architecture. Is it an agent? Is it a workflow? Are you using tools? How do you define your tools? What do your evals look like? If some of these words don't even mean anything to you yet, come check it out. Like it's the easiest way to just learn about how AI products are being built. We release episodes every Thursday. It's called Just Now Possible. It's on Spotify. It's on Apple Podcasts. It's on YouTube and it's on producttalk.org. Awesome. I'm going to go check it out. Thank you so much for being here today. This has been amazing. Thanks for having me.

Teresa Torres is a world-renowned author, speaker, and product discovery coach. Her mission is to help product teams make better decisions about what to build. Teresa uses her Opportunity Solution Tree model as a visual representation of how to identify clear business outcomes; uncover unmet customer needs, pain points, and desires; and discover the right solutions that deliver both business value and customer value.

Her book Continuous Discovery Habits is a considered a must read, and has sold over 140,000 copies. Nearly 20,000 students from around the world have learned from Teresa through the Product Talk Academy, which also reaches hundreds of thousands of product people through free blog posts and social media content.

Teresa is a prolific writer and commenter in the product discovery space. In March 2025, she launched Just Now Possible, a weekly podcast that explores everything from workflows and agents to RAG (retrieval-augmented generation) and evaluation strategies, and digs into how their products keep evolving. And in January 2026, she and Product Talk launched a new course, entitled Business Fundamentals.